2 ms·OpenAI's idea to increase safety is to have AI models police each other1 points by thunderbong 2y ago